By Sam Schechner

The United Nations's chief wants humanity to forbid killer robots.

U.N. Secretary-General António Guterres called Monday for lethal autonomous weapons to be "banned by international law," resurfacing an AI-safety dispute that was core to Anthropic's schism with the Pentagon earlier this year.

"Machines selecting and engaging their target and taking a life — without human control and judgment. That is morally repugnant," Guterres said in a speech about AI governance in Geneva. "Let us call them what they are: killer robots."

Artificial-intelligence systems and chips designed for civilian use are increasingly being deployed both on the battlefield and in military headquarters. That is setting up a debate over when military forces should use AI, when humans should intercede and who should decide where to draw the lines.

The debate burst into public early this year when Anthropic sought assurances from the Pentagon that its AI models wouldn't be used to power either domestic-surveillance or autonomous-weapons systems. The Pentagon, which was then a heavy user of Anthropic tools, refused, saying it should be allowed to use Anthropic's AI for all lawful purposes. The two sides are still fighting in court.

Pope Leo XIV made the case for a ban on AI-controlled weapons and autonomous warfare in his papal encyclical letter this spring, warning that AI systems threatened to " normalize an anti-human vision." He said that AI-driven weapons systems risked lowering the political costs of war for those who possess advanced AI, making "war more 'feasible' and less subject to human control."

But progress is moving quickly on the battlefield. AI-weapons proponents and defense companies argue that autonomous weapons will remain under human control, with the AI allowing human commanders to work with greater — and necessary — precision and speed.

Detractors say that AI systems make mistakes, making it dangerous for humans to cede control to machines in the heat of the moment. Some AI-safety campaigners — many of whom already worry that AI systems will escape human control and turn on their creators — argue that arming AI would only accelerate the danger.

The debate over AI weapons comes as there is growing popular backlash in the U.S. and other countries about the potential negative effects of the AI revolution, including mass job replacement and electricity-price increases.

In the U.S., the Trump administration last month lifted a weekslong export ban on a pair of powerful models from Anthropic, but debate over how to regulate AI systems is still raging. The administration says it is working on a process to test models for safety before release.

Guterres was speaking in Geneva ahead of a week of AI-governance discussions with governments and safety campaigners to discuss ways that AI could be regulated in areas including child safety.

"Let us not wait for atrocity to act," he said. "Some decisions must remain forever human — none more than taking a human life."

Write to Sam Schechner at Sam.Schechner@wsj.com